The LEX AI platform is built on a simple idea: lawyers shouldn't waste time manually searching across dozens of websites. Instead — one question in chat, and the AI finds the right data from every available source.
Today in production we serve 340+ million records from 30+ sources, unified through 64 MCP tools (Model Context Protocol). This article is the complete overview: what we have, where it comes from, and how it works.
The Big Picture
| Category | Records | Tools |
|———-|———|——-|
| EDRSR (court decisions) | ~208M | 6 |
| Court system | 30.5M+ | 7 |
| OpenReyestr + NAIS | 41.8M | 24 |
| Sanctions & anti-corruption | 1.7M | 4 |
| ARMA + Due Diligence | 2M+ | 5 |
| Intellectual property | 295K | 3 |
| Public finance | 1M+ | 4 |
| Verkhovna Rada | 85K | 4 |
| Legislation | 318K | 3 |
| Attorneys & judges | 73K+ | 3 |
| Total | ~340M+ | 64 |
1. EDRSR — The Heart of the Platform (208M Records)
The Unified State Register of Court Decisions is the largest data source on the platform. Two datasets:
- edrsr_documents — 93M metadata records (court, judge, date, category, parties)
- edrsr_fulltext — 115M full decision texts (~1 TB)
What You Can Do
\
"Find Supreme Court decisions on moral damages compensation
for 2024-2025"
\\
The AI selects one of 6 tools:
| Tool | Purpose |
|——|———|
| search_edrsr_decisions\ | Filtered search by metadata |
| search_edrsr_fulltext\ | Full-text search with highlighting |
| search_edrsr_semantic\ | Semantic search by meaning (Voyage AI) |
| get_edrsr_decision_fulltext\ | Full text of a decision |
| get_court_decision\ | Text split into FACTS / REASONING / DECISION |
| get_citation_graph\ | Citation graph between decisions |
Semantic search means you describe a situation in your own words, and the system finds decisions with similar circumstances — even when not a single keyword matches.
2. Court System (30.5M+ Records)
Beyond the decisions themselves, the platform holds data on the entire judicial process:
| Source | Records | Contents |
|——–|———|———-|
| Court sessions | 30.5M | Date, court, judge, parties, outcome |
| Judges (HQCJ) | 417K | Dossiers, tenure, decisions, disciplinary actions |
| Case status | 1.25M | Tracking case movement across instances |
| Session schedule | 480K | Scheduled hearings for 2026 |
| Court experts | 80K | MOJ-certified experts |
| ECHR practice | 11K | European Court of Human Rights decisions |
| HCJ decisions | 16.5K | Disciplinary decisions regarding judges |
| HQCJ (extended) | 4.8K | Qualification, evaluation, vacancies |
| Automatic assignment | 71K | SJAU protocols |
Procedural Tools
A separate group of tools assists with procedural work:
-
calculate_procedural_deadlines\— calculate appeal deadlines by procedural code (CPC, CC, CAS, CrPC) -
search_procedural_norms\— find relevant articles of procedural codes -
build_procedural_checklist\— generate a checklist for a specific case stage
\
"What is the deadline for appealing a commercial court decision?"
→ Article 256 CC: 20 days from the date of the full text
\\
3. OpenReyestr + NAIS (41.8M Records)
11 state registries from data.gov.ua plus EDR data — the most comprehensive database for due diligence:
| Registry | Records |
|———-|———|
| Enforcement proceedings (ASVP) | 29M |
| Debtors registry | 10.4M |
| Individual entrepreneurs (FOP) | 6.9M |
| Company founders | 3M |
| Authorized signatories | 2.8M |
| Legal entities | 2M |
| Notarial special forms | 1.8M |
| Streets (address registry) | 1.5M |
| Administrative-territorial units | 924K |
| Tax debt | 861K |
| Social contribution (SSC) debt | 669K |
| VAT payers | 264K |
| Simplified taxation | 153K |
| Bankruptcy | 36K |
| Notaries | 5.8K |
| Arbitration managers | 3.4K |
| Forensic examination methods | 1.5K |
24 OpenReyestr tools cover: company search, beneficial owners, debtors, enforcement proceedings, bankruptcy, notaries, experts, VAT, SSC, and address data.
Example: Due Diligence in 30 Seconds
\
"Check counterparty by EDRPOU 12345678"
\\
The AI automatically checks:
- Registration in EDR (legal entity / individual entrepreneur)
- Enforcement proceedings (ASVP)
- Debtors registry
- Bankruptcy
- Sanctions lists
- Court decisions (EDRSR)
- Tax debt
The result is a structured report from all sources in a single window.
4. Sanctions & Anti-Corruption (1.7M Records)
| Source | Records | Coverage |
|——–|———|———-|
| OpenSanctions | 1.25M | NSDC, OFAC, EU, UN, UK + 340 programs |
| NAPC declarations | 322K | Official asset declaration checks |
| Corruption registry | 107.5K | Registry of persons involved in corruption |
| Declaration audits | 2K | NAPC audit results |
\
"Is Ivanov Petro Serhiyovych on any sanctions lists?"
→ Search across 1.25M records: NSDC, OFAC, EU, UN, UK, and 340+ other programs
→ Fuzzy matching by name, TIN, passport, EDRPOU
\\
5. Intellectual Property (295K Records)
| Source | Records |
|——–|———|
| Patents (Ukrpatent) | 118K |
| Trademarks | 176K |
| Shareholders (NSSMC) | 1.3K |
Search by name, owner, NICE class (for trademarks) or IPC (for patents), application number.
\
"Find trademarks containing 'Legal' in class 42"
→ 3 results: LEX AI (certificate No. 345678), LegalTech Pro...
\\
6. Public Finance (1M+ Records)
| Source | Records |
|——–|———|
| Prozorro tenders | 1M |
| Spending.gov.ua contracts | 2.8K |
| SSSU financial data | 8.4K |
| Inspection plans | 32K |
7. Verkhovna Rada (85K Records)
4 tools for monitoring parliamentary activity:
| Data | Records |
|——|———|
| Bills | 14.8K |
| Votes | 21.9K |
| Deputies | 463 |
| Deputies' assistants | 4.4K |
| Full legislative texts | 44K |
\
"Which deputies voted for bill 1234?"
→ Full list broken down by faction
\\
8. Legislation (318K Records)
| Source | Records |
|——–|———|
| EDRNPA (cards) | 141K |
| EDRNPA (texts) | 141K |
| Law sections (chunks) | 25K |
| Articles (structured) | 12K |
3 tools for working with legislation:
-
search_legislation\— semantic search across legislative texts -
get_legislation_article\— specific article ("Art. 625 CC") -
get_legislation_history\— amendment and revision history
The system understands aliases: "Constitution", "CC" (Civil Code), "CrPC" (Criminal Procedure Code), "CommC" (Commercial Code), etc.
9. Analytical Tools
Beyond search, the platform includes tools for legal analysis:
| Tool | What It Does |
|——|————-|
| analyze_case_pattern\ | Analyzes arguments, risks, and outcome statistics |
| compare_practice_pro_contra\ | Compares case law "for" and "against" a thesis |
| find_similar_reasoning\ | Finds decisions with similar reasoning sections |
| check_precedent_status\ | Checks whether a precedent is valid / overturned / limited |
| validate_response\ | Anti-hallucination verification of AI responses |
Architecture: How It Works
\
Lawyer → Chat → AI Model → Intent Classifier
↓
Tool Selection (1-5 out of 64)
↓
PostgreSQL / Qdrant / Redis
↓
Structured Response
\\
Each tool is an MCP tool (Model Context Protocol). The AI model autonomously selects which tools to call based on the query context.
Three transports:
- MCP stdio — for Claude Desktop
- HTTP API — for web applications
- SSE — for streaming results
What's Next
Coming up:
- Completing UIPV import — trademarks (46% loaded), utility models (162K), industrial designs (48K)
- DRRP (real estate registry) — agreement with NAIS
- DRORM (movable property encumbrances) — agreement with NAIS
- SLC (State Land Cadastre) — agreement with the State Geocadastre
- Spending.gov.ua — acts, supplementary agreements, penalties (API ready)
- Bulk download RTF — full texts of EDRSR decisions
Summary
LEX AI is more than search. It's a single access point to all of Ukraine's open legal data:
- 340M+ records from 30+ sources
- 64 MCP tools for search, analysis, and verification
- Semantic search — describe the situation, find the decisions
- Due diligence — counterparty check in 30 seconds
- Procedural calculators — deadlines, checklists, norms
All of this is live right now at legal.org.ua.
Register: legal.org.ua
Originally published on legal.org.ua.
Top comments (0)